Novel Process Windows for the safe and continuous synthesis of tert.-butyl peroxypivalate with micro process technology

Based on the economy of scale, the classical chemical industry uses large scale reactors to increase production output and to decrease the average unit costs. This results in large footprint plants consuming land and a huge amount of resources. This large scale character bears certain risks for the environment, health and safety. The industry spends a lot of money for process control techniques to ensure safety and to prevent environmental threats. Hence, alternative production routes or technologies to achieve benefits in above mentioned topics are desirable. For the special case of the organic peroxide synthesis discussed in this thesis, these large scale features bears a very high potential of damage in case of a malfunction. Furthermore, a large amount of cooling energy is required to maintain the exothermic process at a moderate reaction temperature to avoid a thermal runaway. Thus, an intrinsically fast reaction is deliberately slowed down for safety reasons. Based on the overall peroxide process, the use of large scale vessels for storage, downstream processing etc. cannot totally be avoided. But it should be taken under consideration to upgrade or to substitute critical parts of such processes by processtailored microreactors. This can lead to a significant increase in process safety and, in the best case, to an increase in productivity due to the better control of the reaction conditions as e.g. residence time and reaction temperature. The aim of the research described in this thesis was to develop a microreactor for the safe and continuous production of tert.-butyl peroxypivalate. In this thesis the two staged, exothermic and biphasic synthesis of tert.-butyl peroxypivalate is the central chemical process. The first reaction step, the deprotonation of tert.-butylhydroperoxide to potassium tert.-butylperoxide is a slightly exothermic single phase reaction and is conventionally processed at reaction temperatures around 20 °C in batch or in semi-continuous mode. The second reaction step, the formation of the peroxyester by reaction of the potassium tert.-butylperoxide with pivaloylchloride is a two phase reaction and is much more challenging because of the temperature sensitivity of the product and also because of the strong corrosivity of the acid halide. The industrial semi-continuous process bears several safety risks, and it has therefore been recommended to transfer this process into a microreactor supported one. Such a development is always accompanied by the use of different analytical methods. Those used in the context of this thesis are described in Chapter 2. The most commonly used method to determine the amount of organic peroxides is the iodometric titration. Because of the different reactivity of the different peroxides towards the iodide ion, it is necessary to adapt literature-known methods to the special needs of the peroxide which is analyzed. Two reliable iodometric procedures are successfully developed - one for the determination of tert.-butylhydroperoxide and the other one for the determination of tert.-butyl peroxypivalate as single components. Because these methods are time consuming, and only a complex titration procedure allows the simultaneous determination of both organic peroxides present in one sample, it has been decided to use these methods as references for the developed RP-HPLC procedure. This technique allows the fast determination of the reactants and of the product in one step without any complex and time consuming sample pretreatment. The interfacial and the surface tension as well as the wetting angle are parameters that are needed to understand different observations during the slug flow experiments which are discussed in Chapter 3. Therefore, a simple handmade measuring cell for the determination of interfacial and surface tension is presented and the method used to determine the wetting angle is briefly described. Droplet size distributions and residence time distributions are measured for the microreactor characterization (Chapter 7). The temperature profile on top of the reaction plate under reactive conditions using thermal imaging is determined as well. Therefore, these methods are also described in Chapter 2. The possibility to use slug flow for the second reaction step – the conversion of potassium tert.-butylperoxide with pivaloylchloride – is investigated in Chapter 3. This undertaking is very challenging because of the very low interfacial tension of about 4·10-3 Nm-1 and because of the high reactivity of the investigated system. It has been envisaged to gain information on the dependency of conversion and yield on the slug size and on the slug flow velocity. The complexity of the perester system made it necessary to evaluate different types of T-mixers and one Y-mixer for the generation of this highly ordered flow pattern. Investigations are done with the original reactant composition as it is reported for the industrial process and also with a highly diluted system which is needed to increase the interfacial tension. It is shown that an increase in slug flow velocity and a slug size reduction are advantageous for the reaction. This trend is analog to different systems that are reported in literature. Furthermore, it is shown that We-numbers in the range of 10-4 and Ca-numbers in the range of 10-5 are needed to achieve a stable slug formation. Additionally, parameters like wettability and interfacial tension are identified to be of major importance for the generation of a stable slug formation. Contactless conductivity measurements have shown to be suited for the online measurement of slug lengths. The transfer of the tert.-butylhydroperoxide deprotonation into a continuous process is discussed in Chapter 4. This step is a slightly exothermic single phase reaction and is conventionally controlled by step-wise dosing of the reactant. This is a major difference compared to the continuous processing in which the total stoichiometric amount is added in one portion. The latter one is leading to an instantaneous heat release at the point of dosage. It is shown that a shift from batch processing to continuous processing is possible without a loss of process performance. Additionally, it is demonstrated that the potassium tert.-butylperoxide reaction can be done at a very short residence time of 0.4 s. So the length of the residence time section is not determined by the reaction itself but rather by the surface area that is needed for heat removal. Furthermore, no negative influence on reaction performance can be observed until a reaction temperature of 50 °C. This is important for the following chapter which discusses applicability of the Novel Process Window concept to the perester process. The transfer of the second reaction step into a continuous process is discussed in Chapter 5. This reaction step is strongly corrosive, exothermic and biphasic and thus, places special demands on the reactor concept which is used for the continuous processing. First experiments point out that it is necessary to implement static mixing segments into the reaction channel to avoid fast coalescence of the reaction mixture and to keep up the mixing along the reaction channel. Out of the large portfolio of continuous emulsification devices e.g. rotor stator systems, ultrasonic flow cells, membrane emulsifiers, and high pressure homogenizers, the baffle type orifice is chosen because it is the easiest, most flexible and the most straight forward concept that can be realized in a microreactor design. During the process development, the influence of the number of the orifices and that of the spacing between these orifices is investigated. It is shown that the overall process performance is mainly controlled by the number of orifices and that in this specific case, a number larger than five does not contribute significantly to an increase in yield. Furthermore, it is demonstrated that the spacing – as long as it is below a certain length – only contributes to a minor amount of the yield obtained. The shortening of orifice spacing allows a reduction of the residence time by obtaining similar conversion and yield, e.g. using a set-up with four orifices, a spacing of 10 cm and a residence time of 1.2 s gives a yield of 74 %. In comparison to this, a similar yield of 68 % is obtained by a set-up with a spacing of 52 cm and a residence time of 7.6 s. The shortening of spacing, and thus a shortening of the length of the reaction channel, has the effect that the heat cannot be removed sufficiently fast. Hence, the orifice spacing is adapted to the length which is expected to be sufficiently long to remove the heat that is produced after/at each orifice. This leads to the concept of adapted channel length which is then realized in the final orifice microreactor. During these experimental investigations, three Novel Process Windows are identified: ¿ It is shown that it is possible to run a temperature sensitive process beyond the conventional limit which is predetermined by the conventional equipment. The temperature screening up to reaction temperatures of 70 °C shows that the perester reaction performs best at a reaction temperature of 40 °C. Furthermore, it is shown that a microreactor supported process can handle higher temperature fluctuations as compared to the industrial process. This leads to an increase in process safety. ¿ Transport intensification by enhanced mixing results in a significant shortening of process time by achieving yields close to those obtained in the industrial semi-continuous process. ¿ Process simplification is achieved because no step-wise dosing of the acid halide is necessary. Such a processing is not possible in large volume batch reactors. Target of Chapter 6 is to get insights in the kinetic of the second step of the tert.-butyl peroxypivalate synthesis because no data is publicly available. The estimation of the reaction order is done on the basis of the integral method. The experimental data is taken from two set-ups by which pseudo-homogeneous reaction conditions are assumed. Because of the high exothermicity, it is not possible to assure isothermal conditions during these measurements and thus, the determined values can only be used as a rough guide on the kinetics of the tert.-butyl peroxypivalate formation. It is found that the reaction can be described by a 2nd order reaction kinetic. This is confirmed by comparable reactions found in literature. The estimated energy of activation using the data from a conceptual design set-up with nine orifices and a spacing of 5 cm is 3,085 ± 1,97 kJmol-1 the corresponding pre-exponential factor is 2.8 ± 0.2 Lmol-1s-1. The energy of activation, which is estimated from the data using a conceptual design set-up with five orifices and a spacing of 10 cm is 18,202 ± 548 kJmol-1. A corresponding preexponential factor of 1588 ± 98 Lmol-1s-1 is determined. This data in combination with a first order reaction kinetic describing a general decomposition reaction of each of the reactants is then used in a basic plug-flow-tubular reactor model. Both set-ups are considered separately because of the strong variation of the kinetic parameters. The calculated concentrations are compared with the experimentally determined ones. Another method that is discussed is the use of a global rate constant to describe the product formation. This global rate constant in combination with the kinetic parameters that are describing the general reactant decomposition is then used to describe the course of reaction. The obtained global reaction rate is 1.32 Lmol-1s-1 using the data of the conceptual design set-up with nine orifices and a spacing of 5 cm. That obtained with the data from a conceptual design set-up with five orifices and a spacing of 10 cm is 1.25 Lmol-1s-1. All kinetic parameters should be the same for both set-ups, since they are not reactor dependent. The broad variation of these kinetic parameters is caused by experimental errors that occurred during their determination. On the basis of the knowledge gained during this research, an orifice microreactor for the continuous production of tert.-butyl peroxypivalate is developed and manufactured. The meandered reaction channel is made half on one side of the reaction plate and the other half on the other side of the reaction plate. Both reaction plates laid on top of another form the reaction channel including the two caterpillar micromixers and the five orifices. These two plates are then covered in a casting. The heat exchanger is fabricated in the bottom of the casting and is located below the reaction plate. A window that enables infra-red measurements on top of the reaction plate is embedded in the top part of the casting. In Chapter 7 its non-reactive and reactive characterization is discussed and constructional considerations are explained. The conceptual design set-ups on which the microreactor development is based use baffle type orifices whereas the orifices that are installed in the orifice microreactor are of conical type. The change towards the conical design is done to reduce the energy dissipation and to use the flow energy more efficiently for the droplet disruption. Droplet size distribution measurements are made to compare both orifice types with each other. As an example, a mean droplet diameter of 2.8 µm at a flow rate of 75 mLmin-1 resulting in an energy density of 2705 kJm-3 is obtained with the baffle type system. In comparison to this, a mean droplet diameter of 6.6 µm at an energy density of 1231 kJm-3 is obtained with the orifice microreactor using the same flow rate. Hence, the change in geometry towards the conical design reduced the energy dissipation but influenced the droplet size. The different behavior of both types of orifices at comparable flow rates is explained with the help of computational fluid dynamic (CFD) simulations. It is shown that the gradient of the strain rate is higher for the baffle type system as it is for the conical one. Thus, the positive effect of the conical design on the reduction of energy dissipation is superimposed by the influence of reduced strain rate. There is evidence that the positive effect of the conical design on the reduction of energy dissipation plays a more important role at higher flow rates. Considering the issue of scale-up, the reduction of dissipation losses is an important benefit since the throughput can be increased by keeping the homogenizing pressure as low as for the baffle type system at lower flow rates. An additional parameter that is used to describe the quality of an emulsion is the SPAN. This parameter describes the droplet size variation within the formed emulsion. The smaller this parameter, the less variation is there between the sizes of the droplets in the formed emulsion. Considering the data given for the baffle type system at a flow rate of 75 mLmin-1 the SPAN is 1.8 whereas that for the conical type is 1.1. This indicates the positive influence of the change of the orifice geometry on the droplet size variation within the formed emulsion. Based on the theory of droplet disruption in laminar and turbulent flow, one equation is found which can be used to estimate the maximum droplet size using the developed reactor. For the baffle type system three correlations are found that fit to the experimental data. To address the safety aspect, the Frank Kamenetskii approach is used to estimate the critical channel diameter at which a thermal runaway can be excluded. Depending on the assumed energy of activation, a critical channel diameter ranging from 246 µm to 1418 µm can be estimated. Thus the orifice microreactor reaction channel (d = 710 µm) is within these critical diameters. According to that, a thermal runaway cannot totally be excluded but is unlikely. The velocity field in the meandered reaction channel is calculated with the help of CFD simulations. It is shown that the meander design of the reaction channel induces secondary vortices which are indicated by a high Dean number. For example for a flow rate of 10.5 mLmin-1 a Dean number of 89 is obtained and for a flow rate of 21 mLmin-1 a Dean number of 179 is found. In addition to the high overall heat transfer coefficient of 2358 WK-1m-2 this results in a very good heat management of the developed microreactor. Infra-red measurements under reactive conditions are used to confirm this behavior. Therefore, the temperature profile is measured on top of a thin plate covering the reaction channel. The influence of utility fluid flow rate on the reaction performance is demonstrated for two different reaction media flow rates. With these measurements it is shown that the tert.-butyl peroxypivalate process performs best at a reaction temperature of 40 °C. This is 20 K higher as the normal operation mode as it is defined for the industrial semi continuous process. Hence, the Novel Process Window aspects which are identified in Chapter 5 are confirmed. The accuracy of the manufacturing process is checked by residence time distribution and by laser profilometer measurements. Nearly ideal plug flow behavior is indicated by very high Bodenstein numbers of 158 and 180 and by cell numbers of 79 and 90. Also the theoretical residence times of 3 s at a flow rate of 10 mLmin-1 and 1.5 s at a flow rate of 21 mLmin-1 are experimentally confirmed. A benchmark, calculated on the basis of the product output of the industrial process, illustrates the potential of micro process technology for the intensification of industrial processes. By the example of four fictive large scale orifice microreactor processes, it is shown that the footprint as well as the reaction volume can significantly be reduced. The space-time-yield for the industrial process is 0.187 kgL-1h-1 whereas that for the orifice microreactor process, done at 40 °C, is 414 kgL-1h-1. The large scale microreactor has a reaction volume of 0.48 L and is by a factor of 2200 smaller than that for the industrial process.